OpenAI wants to enter the shopping space in a more meaningful way.
The technology giant announced Monday that it has launched new capabilities to help users shop and compare items more efficiently.
More from Sourcing Journal
According to March data from Elon University, 52 percent of U.S. adults use large-language models (LLMs), and among those who have used and LLM, nearly three-quarters have experimented with ChatGPT. Prior to this announcement, OpenAI saw an influx of users turning to ChatGPT for shopping inspiration , said Saguna Goel, ChatGPT shopping product lead at OpenAI.
“This is a use case that we’ve seen users already come to ChatGPT for. We saw a really big spike around this for the holiday season , because ChatGPT is just starting to become this destination for people to come do product research,” Goel said.
But when users previously went to OpenAI for help with shopping, Goel said she and her team identified various roadblocks they might hit on that journey; she noted ChatGPT was often incapable of comparing product reviews and pricing or providing accurate links for consumers to purchase suggested items.
“While we were serving some results on, ‘Here’s what could be interesting for you,’ there wasn’t actually a great next step,” Goel said.
She hopes the latest iteration of ChatGPT—which includes additional features besides shopping—will help consumers change that. The new shopping function will, at first, hone in on categories like apparel, beauty, home goods and electronics.
The main difference between ChatGPT’s shopping function and a cursory Google search for products is that consumers can use natural-language prompts to speak to ChatGPT, rather than using search terms. So, for instance, while a consumer might use Google to search, “linen vacation outfit,” they can instead give ChatGPT a more specific prompt, like, “I’m going to Florida on a girls’ trip in June. I want to purchase a matching set, preferably made of linen, for the trip. My budget is $60.”
From there, the system offers up several different options that meet the specifications the user has outlined in their query. The shopping function can aggregate customer reviews and sentiment from social media sites like Reddit; can provide comparative analysis on price or features and can share direct links to help users make a purchase.
The updated ChatGPT also incorporates product photos into its responses, so that users can shop visually, just as they would do based on an online search. Goel said she believes that function will resonate with consumers, and noted that OpenAI plans to further improve the visual aspect of the shopping function over time.
“We know that users are very visual and aesthetic, especially when it comes to shopping for certain categories, and being able to bring our most intelligent multimodal capabilities at OpenAI and combine that with shopping intent and shopping journeys is something I’m really, really excited about,” Goel said.
OpenAI has started rolling out the updated features to free, Pro and Plus users—logged in or logged out—using the search function on ChatGPT-4o, its latest, most intelligent model. The full release will occur over the next several days. Goel said, for the moment, the upgrades are targeted at consumers, which means that enterprise clients’ systems will not be impacted.
Today, companies cannot purchase paid placements or advertisements for products in ChatGPT’s user-facing results, Goel said. At present, OpenAI does not use affiliate links for products, nor does it generate revenue from product recommendations. Brands and retailers’ offerings are ingested by OpenAI’s technology so long as they aren’t opted out of its search crawler, OAI-SearchBot.
That helps ensure users’ experience with ChatGPT remains as organic as possible, Goel noted.
“We think the best way to build a shopping experience is [to] have ChatGPT just serve you contextual and best recommendations for your query,” she said. “Over time, users might have preferences on specific brands, which they can tell us about, but at this time, none of the recommendations that you see on ChatGPT are going to be…sponsored.”
To that end, OpenAI plans to use existing user data to foster personalized results for users discovering products on ChatGPT. In the coming weeks, the company will launch a memory integration function, which will allow the generative AI tool to recall specific details about a consumer based on past conversations, if a user has ChatGPT’s memory function turned on. Using that information, ChatGPT will soon be able to take preferences into account when making specific product recommendations. For instance, if ChatGPT’s memory knows a user prefers to wear neutrals with a pop of color, it will be able to suggest items based on that knowledge.
Goel said the company plans to continue iterating on the tool as it learns from users’ preferences and actions, but noted that her team looks forward to helping transform users’ shopping experiences in a new way.
“ChatGPT Search has already, from a technology and product perspective, made a really big dent, and I’m excited to…bring [to users] the model that we’ve developed, specifically trained for shopping, so that it can help users get to a place where they’re like, ‘Okay, shopping is not that overwhelming anymore, and I actually have an assistant that is on my side,'” Goel said.
